Using the firm-level census data, this paper re-estimated the total factor productivity (TFP) of firms in China's iron and steel industry and examined its potential determinants over the period 1998-2007. To deal with the "endogenous input" problem, we used the semi-parametric regression techniques for estimating the firm-level TFP. The results suggest that firms' TFP in China's iron and steel industry has been steadily increasing over time with the key drivers of productivity improvement differing substantially between firms with different characteristics including their size, ownership type and geographical location. Notably, the productivity of small firms is positively related to market share and negatively related to R&D. Large state-owned enterprises' productivity is relatively insensitive to changes in market share and R&D, while the non-state owned enterprises are more likely to obtain their productivity gains through export. Increasing firm size is generally positively correlated to firms' performance in TFP, and it is more so in the less developed Western than the Eastern and Central regions. The findings suggest that different policy instruments targeting firms with different characteristics in the process of restructuring the industry may be desirable.